I did point out that they created precisely the element you indicated was challenging: accomodating different document types from different sources.

Given that the USPTO's system is likely based on electronic filings at this point, even the scan-and-present aspects of the BookReader solution are likely more than is required.

Another salient point is that by breaking the problem into discrete components, you're more likely to come up with a viable solution, or at least significant portions of one.

And I've worked on a number of systems (FDA clinical information systems) which address a great deal of your other feature points. Suffice to say: such systems do in fact exist.

They do, in fairness, take more than a couple of days to assemble.